Bioinformatic Prediction of SNPs within miRNA Binding Sites of Inflammatory Genes Associated with Gastric Cancer

Polymorphisms in miRNA binding sites have been shown to affect miRNA binding to target genes, resulting in differential mRNA and protein expression and susceptibility to common diseases. Our purpose was to predict SNPs (single nucleotide polymorphisms) within miRNA binding sites of inflammatory genes in relation to gastric cancer. A complete list of SNPs in the 3’UTR regions of all inflammatory genes associated with gastric cancer was obtained from Pubmed. miRNA target prediction databases (MirSNP, Targetscan Human 6.2, PolymiRTS 3.0, miRNASNP 2.0, and Patrocles) were used to predict miRNA target sites. There were 99 SNPs with MAF>0.05 within the miRNA binding sites of 41 genes among 72 inflammation-related genes associated with gastric cancer. NF-κB and JAK-STAT are the two most important signaling pathways. 47 SNPs of 25 genes with 95 miRNAs were predicted. CCL2 and IL1F5 were found to be the shared target genes of hsa-miRNA-624-3p. Bioinformatic methods could identify a set of SNPs within miRNA binding sites of inflammatory genes, and provide data and direction for subsequent functional verification research.


Introduction
Gastric cancer is one of the most malicious diseases of the world, and the second most frequent cause of cancer deaths (Hartgrink et al., 2009).Chronic inflammation is a very important factor for gastric cancer, and contributes to about 25% of all gastric cancer cases worldwide (Hussain et al., 2007).It is well established that Helicobacter pylori (Hp) associated chronic gastritis will lead to gastric cancer going through a classic process of "chronic gastritis, intestinal metaplasia, dysphasia, gastric cancer" (Konturek et al., 2009).Hp associated Chronic gastritis will induce high expression of some inflammatory cytokines; studies have verified that inflammation and inflammatory cytokine genes play a very important role in the oncogenesis of gastric cancer (Grivennikov et al., 2010;Schetter et al., 2010).Chemokines are combined with its ligands and receptors, which are downstream of pro-inflammatory cytokines, and the components of the chemokine system can recruit leukocyte, cause neo-angiogenesis, and promote tumor cell growth, proliferation and survival, invasion and metastasis, such as CCL2, CC12, CXCR4 (Gonda et al., 2009;Allavena et al., 2011;Wu et al., 2013).
Bioinformatic and cloning studies have estimated that miRNAs may regulate 30% of all human genes (Lewis et al., 2003).miRNAs can regulate genes by pairing to the 3' untranslated regions (UTRs) of messenger
In the previous genetic studies, researchers have placed more interest on the gene function region, such as gene coding region and promoter region.However, the introns and proximal untranslated regions remain unexplored.The discovery of miRNA makes the study of these untranslated regions an important research (Lheureux et al., 2011).In this study, we will just focus on the miRNA binding sites SNPs located in the 3'UTR of inflammation-related genes by using bioinformatic methods.It will provide data for the follow-up studies and functional verification tests, and build evidence for diagnosis and treatment of gastric cancer.

Analysis of candidate inflammation-related genes and their pathways
Published studies that focused on inflammation-

Search of the 3'UTR polymorphisms of candidate genes
The "database SNP" (dbSNP 138) (http://www.ncbi.nlm.nih.gov/SNP/) was used to have a complete list of SNPs within those genes, and the SNPs in 3'UTR region were selected.The alleles and their frequencies (HapMap-HCB, HapMap-Human China Beijing) of the SNPs were read carefully, and the SNPs with MAF (minor allele frequency) higher than 0.05 in HCB were chosen and recorded.

Computational predictions of miRNA target binding sites
Putative miRNA-binding sites within the 3'UTR SNPs of each inflammation-related gene associated with gastric cancer selected above were identified by means of specialized miRNA target prediction databases: MirSNP (Liu et al., 2012)    element) sequences in the 3'UTR of target mRNAs, while miRNA sequences were got from miRBase 18 (http:// mirbase.org).miRNA's function was affected by the SNP in the 3'UTR of target gene.SNPs within the target sites could analogously modulate the miRNA-mRNA interaction and decrease, break, enhance and create a miRNA-mRNA binding site (Liu et al., 2012).'Create' means that when the allele changes from wild to variant, a new miRNA binding site is created.'Break' means the original miRNA binding site is broken.'Enhance' means the ability of miRNA and target gene binding is enhanced, presenting the miRNA binding site with one more base-pair.Meanwhile, 'decrease' means the miRNA binding site was reduced by one base-pair.The function was directly got from MirSNP or by miRNA binding site of the miRNA and SNP sequence.

Assessment of the binding free energy
The sequences of SNPs and miRNAs were respectively got from the database of dbSNP and miRBase.
RNAcofold (http://rna.tbi.univie.ac.at/cgi-bin/RNAcofold.cgi),this database was used to assess the Gibbs binding free energy (rG, expressed in kJ/mol) for the wild and the variant alleles, then the difference of the free energies between the two alleles was computed as "variation of rG" (i.e., rrG) (Landi et al., 2011).

Network analysis of the interaction between miRNAs and mRNAs
Cytoscape software (version 2.8.3,National Institute of General Medical Sciences (NIGMS), U.S.) was used to visualize the network of the target genes and the related miRNAs, and terms with attribute of interest were highlighted (Ross et al., 2013;Song et al., 2013;Spinelli et al., 2013).Specifically, we build a network centered on the inflammation-related genes and the corresponding miRNAs.In the network, each node was an entry, and two nodes were linked by an edge if they had a relation by the above bioinformatic prediction.Node or edge attributes represented entity descriptions and relations annotations.

The inflammatory genes associated with gastric cancer
After searched the data library of PubMed, Web of knowledge, and Ovid, schematic (Figure 1).representation the mechanisms for the involvement of inflammation and inflammatory genes in gastric cancer development.NF-κB and JAK-STAT are the two most important signaling pathways.72inflammation-related genes (Table 1) were found to be related to gastric cancer.They were classified into pro-inflammatory cytokines, receptors, and chemokine genes, anti-inflammatory cytokines, receptors, prostaglandins and nitric oxide.Among these genes, there were 9 anti-inflammatory cytokines, receptors (including IL4, IL5, IL5RA, IL10, IL10RA, IL10RB, IL13, IL13RA1, and TGFB1) and 2 prostaglandins and nitric oxide (COX2 and INOS).Other 61 genes were proinflammatory cytokines, receptors, and chemokine genes.

SNPs' information of these inflammatory genes
Among these genes, SNPs without data of HapMap-HCB, or with MAF ≤ 0.05 in HapMap-HCB were all excluded.99 SNPs of 41 genes in the 3'UTR with MAF>0.05 were selected.There are 35 pro-inflammatory genes with 81 SNPs, and 6 anti-inflammatory genes with 18 SNPs (Table 2).And there are more than one SNP of some genes, such as CCR2, IL1F5, and IL4R.

The results of target SNPs predicted by each miRNA target database
After 99 SNPs with HCB MAF>0.05 locating in 3'UTR of these 41 genes were examined, 47 SNPs of 25 inflammatory genes were found to have the miRNA binding sites.Among these genes, there were 21 proinflammatory genes and 4 anti-inflammatory genes.
And there were 95 putative miRNAs.85 miRNAs were obtained by searching in MirSNP, 11 miRNAs were obtained by searching in TargetScan Human 6.2, 25 miRNAs were obtained by searching in PolymiRTS 3.0, 24 miRNAs were obtained by searching in miRNASNP 2.0, and 5 miRNAs were obtained by searching in Patrocles (Table 3).
The functions of the SNPs in the 3'UTR of inflammatory genes to miRNAs were represented in Table 3.

The network of target genes and miRNA interaction
There were 2 disjoint sets of nodes in this graph, mRNA genes (gray circle) and miRNA (white circle).A direct connection placed from a miRNA to an mRNA indicates that the mRNA was predicted to be the target of the miRNA.The length of the line between miRNA and target gene indicated the size of SNPs affecting miRNAs.The shorter the line was, the bigger the effect was.The resulting network had 95 white nodes and 25 gray nodes.CCL2 and IL1F5 were found to have the shared target gene of hsa-miRNA-624-3p.

Discussion
Genetic changes in regulatory regions have been reported to play an important role in susceptibility of common diseases.Variants in 3'UTR of some genes were reported to have relation with higher susceptibility to certain type of cancers (Lheureux et al., 2011;Wang et al., 2012;Iuliano et al., 2013;Skeeles et al., 2013).SNPs within the miRNA binding sites have been shown to affect the ability of miRNAs and target genes binding, resulting in abnormal mRNA and protein expression (Kertesz et al., 2007;Wang et al., 2008;Skeeles et al., 2013), which will influence the susceptibility risk of certain cancers.Compared with previous genetic studies which focused on the gene coding region, it's also very important to identify a set of SNPs within miRNA binding sites of the inflammatory genes associated with gastric cancer for cancer research.
In consideration of cost-effectiveness, a selection criterion is made in this study that the MAF of the SNPs is higher than 0.05, otherwise it will be excluded (Engels et al., 2007).In this study, although lots of SNPs were searched out in the 3'UTR of 72 inflammatory genes which were found to have association with gastric cancer, at last, most SNPs without data of HapMap-HCB or with MAF≤0.05 were excluded, and there were only 99 SNPs in the 3'UTR of 41 genes selected.Therefore, it's necessary to carry out case-control studies which can provide accurate frequency data of alleles.Then, more susceptibility SNPs will be found out.However, the inferred target SNPs predicted by miRNA target prediction databases were just theoretical results.More evidence from case-control studies and luciferase assays are needed to further validate whether they are the real functional sites.
In general, miRNA could form an actively sTable Watson-Crick base pair with its target mRNA (Bartel 2009).In most occasions, the seed sequence is located at the position 2 to 8 from the 5' end of the miRNA, and acts as an essential scaffold for recognizing the target mRNA by matching with MRE sequences of mRNA (Satoh 2012).Because of thermodynamic rule and the evolutional conservation of MRE sequences, it's possible to accurately predict target mRNAs of miRNAs by computational approaches comparatively.miRNA carries out its function by binding to target gene, so it is crucial to identify its target gene.In recent years, more and more databases have been used to explore the impact of SNPs on miRNA binding sites and are open to public, which respectively are MirSNP, TargetScan Human 6.2, miRNASNP 2.0, Patrocles, and PolymiRTS Database 3.0.They are based on different algorithms, such as sequence complementarily between miRNA and its target gene and the binding energy of the miRNA-target double-stranded, and they will give different predicted results.Therefore, this study listed out all the target SNPs that predicted by these five miRNA target prediction databases.The more databases predicted the SNP, the more likely it would be the true target SNP.
Compared with four other existing databases (TargetScan 6.2, miRNASNP 2.0, PolymiRTS 3.0, and Patrocles), MirSNP prediction was most sensitive.47 miRNA-related SNPs were identified, and also 85 related miRNAs, which accounted for most of the results.MirSNP is based on information from mirBASE18 and dbSNP135, and it has been developed to identify putative miRNA-related SNPs and miRNAs from single data sets of GWAS (genome-wide association study) or eQTL (expression Quantitative Trait Loci), especially from the newly published datasets (Chenxing Liu et al., 2012).A SNP within the target site could decrease, break, enhance and create a miRNA-mRNA binding site, thus affecting the function of miRNA (Ryan et al., 2010;Liu et al., 2012).A large number of records of SNPs within predicted miRNA target sites are stored in the MirSNP database (Chenxing Liu et al., 2012), and the effects of SNPs on miRNAs could be got directly from MirSNP.So it provides a convenient search platform.
RNAcofold is one of the core programs of the Vienna RNA package (http://www.tbi.univie.ac.at/~ivo/RNA/), which can be used to predict the hybridization energy and base-pairing pattern of two RNA sequences (Gruber et al., 2008;Landi et al., 2011).It is based on concatenating the two RNA sequences and treating the loop containing the concatenation point as an exterior loop.Because of the use of Zuker algorithm, some common interaction motifs such as kissing hairpins can not be predicted (Gruber et al., 2008;Landi et al., 2011).
In this study, the RNAcofold was used to get the rG for the wild and the variant allele of the target SNP, and it was computed as rrG for the difference of the free energies between the two alleles.The absolute values of rrG for each miRNA were listed out.It can be used as parameter for predicting the biological impact of each target SNP.The higher absolute value of rrG, the bigger impact SNP on miRNA binding site, and the more product of target gene influenced (Landi et al., 2011;Lipchina et al., 2011).So it is more meaningful for the further experiments of the miRNA function with high absolute value of rrG.However, since the inference was based on rules summarized from current uncompleted published data, some exceptions were possible and more experimental data are needed to validate the results.
miRNA exerts function by pairing to the 3' UTR of target genes which lead to mRNA degradation or repression of protein synthesis (Carthew 2006).Owing to their ability to interact with mRNAs, miRNAs can act as oncogenes or tumor-suppressors, depending on the levels of their expression (He et al., 2005).Several miRNAs have been reported in relation to tumorgenesis, while some have the opposite functions of reducing inflammation and inhibiting malignancy in the inflammation pathway (Zabaleta, 2012).For example, hsa-miR-155 inhibited the production of the pro-inflammatory cytokine IL8 by inhibiting the NF-κB pathway (Crone et al., 2012), and the levels of hsa-miR-204 in the gastric mucosa were significantly increased after H. pylori eradication (Shiotani et al., 2012).While Overexpression of miR-150 promoted the proliferation of gastric cancer cells (Wu et al., 2010).Therefore, the identification of cancer related miRNAs and their target genes in the inflammation pathway are important for gastric cancer biology research and its treatment.
In summary, bioinformatic methods could identify a set of SNPs within miRNA binding sites of the inflammatory genes associated with gastric cancer and the corresponding miRNAs.miRNA function was affected by the SNP in the 3'UTR of target gene.This could provide data and direction for subsequent functional verification researches, minimize the costs and narrow the range of experiments.It is very important for gastric cancer biology research.However, the predicted target SNPs and miRNAs were just theoretical.More case-control association studies and function verification experiments such as luciferase report system are needed to carry out.

Figure
Figure 1.The Pathways from Stomach Inflammation to Gastric Cancer

Figure 2 .
Figure 2. miRNA -mRNA Interaction Network Construction.The inflammation-related genes nodes were shown in gray; and the related miRNAs were shown in white